Significant artefactual noise in (90)Y TOF-PET imaging of low specific activity phantoms arises despite increased acquisition time

Volumes of usual PET phantoms are about four to sixfold that of a human liver. In order to avoid count rate saturation and handling of very high (90)Y activity, reported TOF-PET phantom studies are performed using specific activities lower than those observed in liver radioembolization. However, due to the constant random coincidence rate induced by the natural crystal radioactivity, reduction of (90)Y specific activity in TOF-PET imaging cannot be counterbalanced by increasing the acquisition time. As a result, most (90)Y phantom studies reported images noisier than those obtained in whole-body (18)F-FDG, and thus advised to use dedicated noise control in TOF-PET imaging post (90)Y liver radioembolization. We performed acquisitions of the Jaszczak Deluxe phantom in which the hot rod insert was only partially filled with 2.6 GBq of (90)Y. Standard reconstruction parameters recommended by the manufacturer for whole-body (18)F-FDG PET were used. Low specific activity setups, although exactly compensated by increasing the acquisition time in order to get the same number of detected true coincidences per millilitre, were impacted by significant noise. On the other hand, specific activity and acquisition time setup similar to that used in post (90)Y liver radioembolization provided image quality very close to that of whole-body (18)F-FDG. This result clearly discards the use of low specific activity phantoms intended to TOF-PET reconstruction parameter optimization. Volume reduction of large phantoms can be achieved by vertically setting the phantoms or by adding Styrofoam inserts.